Fall 2016

Abstract

The final project involves analyzing data about basketball players from NBA League in the 2015-2016 season.

The motivation question is :

"In the 2015-2016 season, how do the skills of a player relate to his salary?"

We start from data acquisition and data cleaning

There are three kinds of raw data tables : Roster, Totals, Salaries

Steps we do here:

  1. scrape the three kinds of tables and get 90 raw data files

  2. produce roster-salary-stats.csv file, which contains all variables from Roster,
    Totals and Salary.

  3. write auxiliary cleaning functions

Picture of code we use to scrape and download data

Continue

Exploratory Data Analysis

Two aspects of analysis : Qualitative amd Quantitative

Qualitative variables:player's name, birth date, country, team, position, etc.

Quantitative variables: mean, median, min, standard deviation, etc.

Example of Qualitative analysis


bar chart of counts of positions


bar chart of counts of number of players in each team

Examples of Quantitative analysis

This is part of eda-output.txt file

This histogram shows the frequency of field goals.

We can see that the highest frequency of salary centered around 2000000.

Shiny App for "team-salaries"

This bar chart shows the minimum salary for each team in ascending order. We can see that LAL is the team with highest minimum salary.

This bar chart shows the total payroll for each team in ascending order.

This bar chart shows the average salary for each team in descending order.

This bar chart shows the standard deviation of salary for each team in ascending order.

This bar chart shows the first quantile of salary statistics for each team in descending order

Analyze performace of players by using Eff Model

Old Eff formula = (PTS + REB + AST + STL + BLK - Missed FG - Missed FT - TO) / GP

Only favors offense-oriented players. Not good!

So we use:

New EFF = w1x1/std1 + …w8x8/std8, where xj is the j-th variable, j = 1,2,…,8, wj is the weight for variable j = 1,2,…8, stdj is the standard deviation of variable j = 1,2,…,8

Table "eff-stats-salary.csv", which contains efficiency statistics and other important variables
For position "PF". Same thing for other positions.

Shiny app "stat-salaries"

Positive relation between total points and salary

The picture shows the relation between salary and missed free throws.

There are a few outlier data, which indicate that some players missed a lot of free throws although they get a very high salary.
## Calculate the value of a player

Formula: value = efficiency/salary. We want to identify the top 20 players (most valuable) and the bottom 20 players(worst value)

Here is a picure of most valuable and least valuable players:

Conclusion from our analysis of this project:

From our analysis, although there is a positive relation between players' salary and their efficiency, there are some players who have high salaries but missed a lot of field goals, and some players who earn average or low salaries actually perform well. So basketball teams should use an accurate statistical method such as PCA and modified EFF formula to analyze players' performances as one of standards to determine their salary.